I Ran My Own Writing through AI Detection Software

What happens when care, clarity, and serious prose are mistaken for code?

I ran a piece of my own writing through AI detection software and was told that it was 74% AI.

The number was absurd. The experience was not.

What unsettled me was not only that the result was false. It was that the falsehood touched something culturally revealing. The software did not seem to be reacting to plagiarism, fraud, or even to some obvious imitation of machine prose. It seemed to be reacting to the very qualities I most value in writing: rigor, clarity, composure, structure, conceptual sequence, and stylistic control. In other words, it was not flagging carelessness. It was flagging form.

When disciplined prose begins to look suspicious

That is not only irritating. It is clarifying.

It reveals a confusion that runs deeper than any one faulty tool. We may be entering a period in which disciplined prose is increasingly legible to some readers, and to some systems, as less human rather than more. Writing that bears the marks of revision, proportion, and conscious construction may now appear suspicious precisely because it does not advertise enough visible mess. The problem is not simply that detection software gets things wrong. The problem is that what it gets wrong may reflect a broader loss of confidence in our ability to recognize what human authorship looks like when it has been seriously cultivated.

For many people, the image of human writing remains tied to spontaneity, irregularity, warmth, confession, idiosyncrasy. On that view, what feels immediate feels alive. What feels composed begins to look suspect. This was always a limited understanding of prose, but it becomes more unstable in a world where machine systems can now generate smooth, competent language on demand. Once that happens, surface fluency no longer reassures. Some readers begin searching for other signs of humanity. They look for roughness, hesitation, verbal fingerprints, little disruptions that seem to certify the presence of a person on the page.

What falls out of view in that search is the possibility that one mark of human seriousness is not roughness but achieved form.

A sentence can be carefully made without being mechanical. A paragraph can hold its shape because someone thought through its burden and sequence. An argument can move cleanly because the writer has spent years learning how to place each concept where it belongs. None of that is evidence against humanity. It is evidence of discipline.

This matters especially to those for whom writing is not merely expressive but exacting. For some writers, the placement of a word is not cosmetic. It is part of the attempt to transmit thought faithfully. The difference between one sentence and another is not a matter of decoration. It is the difference between approximation and accuracy, between what merely gestures toward an idea and what actually carries it. Anyone who works this way knows that form is not something added after thought. It is one of the ways thought becomes visible at all.

That is why a false result from a detector can land with more force than it might appear to deserve. The injury is not confined to the number. It touches the labor beneath the prose. It implies that what one has spent years refining may now be read as evidence of impersonality. It can make a serious writer feel, if only for a moment, that the very signs of inward care have become culturally illegible.

That feeling deserves to be named.

Not because every writer needs comfort. Not because the judgment of a bad tool should be granted moral authority. But because for those who take language seriously, such misrecognition is not trivial. It can feel strangely intimate. One does not spend years learning how to think on the page without becoming attached to the integrity of that process. To have the result flattened by a crude instrument can produce not only annoyance but a kind of estrangement. The writer knows what it cost to arrive at that prose. The software sees pattern.

Fluent language and inhabited language

That difference matters.

The deeper issue, then, is not merely whether a detector is accurate. It is whether we still know how to distinguish fluent language from inhabited language.

Fluent language is now widely available. It can be produced rapidly, convincingly, and at scale. It can sound balanced, polished, and informed. It can simulate sequence. It can adopt tone. It can produce the local atmosphere of intelligence. What it cannot guarantee is the presence of a mind that has actually lived among the distinctions it is making, suffered the resistance of its own material, rejected easier formulations, or taken responsibility for what is being said.

That is where inhabited language begins.

Inhabited language is not defined by roughness. Nor is it guaranteed by elegance. It is defined by contact. Something has been seen, weighed, revised, and borne. The words are not merely arranged competently. They are placed under pressure from within. The movement of the prose reflects judgment rather than mere continuation. One feels, even when the style is restrained, that the language answers to something more than probable verbal success.

This is not always easy to describe, and perhaps that is part of the problem. Many of the most important marks of serious writing are not reducible to metrics. They are felt in proportion, moral weight, earned distinction, the refusal of the merely impressive phrase, the sense that a sentence has been made answerable to reality rather than merely to expectation. These are not magical qualities, and they are not infallible signs. But they belong to the old human disciplines of reading and writing, and no detector can perceive them simply by counting surface features.

A detector can identify resemblance. It cannot determine inhabitation.

That limitation would be less interesting if it remained merely technical. What gives it wider significance is the possibility that our own habits of reading may be growing more detector-like. Instead of asking what kind of mind a piece of writing reveals, we may be drifting toward quicker impressions. Does it sound polished. Does it feel intelligent. Does it resemble the style now associated with machine output. Does it have the slightly overcomposed quality people have learned to distrust. These are not worthless perceptions, but they are shallow ones if they become primary.

The danger is not only that people will be falsely accused. The danger is that a culture saturated with generated fluency may slowly lose contact with the difference between having an impression and having formed a thought.

Those are not the same event.

The difference between an impression and a thought

An impression is immediate. It arrives quickly. It may be strong, persuasive, even illuminating. But it does not yet carry the burden of judgment. It has not necessarily been tested, differentiated, or placed in relation to competing realities. A thought, in the fuller sense, has undergone more. It has been worked on. It has survived contact with resistance. It has taken form through the disciplines required to make it more than reaction.

Language can produce the feeling of thought without its formation. This has always been true. Rhetoric, fashion, prestige, and emotional force have always been able to create the appearance of understanding. But the scale and ease of generated language make the problem harder to ignore. When coherent prose can be produced so readily, the old temptation intensifies. It becomes easier to settle for verbal adequacy. Easier to confuse smoothness with substance. Easier to accept the sensation of clarity in place of the harder work of asking what has actually been seen, known, and made.

This is why the episode with the detector feels larger than itself. It is not important because a faulty tool hurt a writer’s feelings. It is important because it stages, in miniature, a cultural failure of discrimination. A prose shaped by care is mistaken for simulation because the surrounding culture has become uncertain about how to recognize formed thought when it appears in composed language.

That uncertainty affects both writers and readers.

The burden now

For writers, it can create a subtle pressure to perform humanness in increasingly obvious ways, to leave seams showing, to cultivate roughness as evidence of authenticity, to distrust one’s own finish. That would be a mistake. A writer should not have to scatter imperfections across the page in order to reassure a suspicious age. Deliberate slackness is no answer to cultural confusion. The task is not to abandon form. It is to defend the difference between form and emptiness.

For readers, the burden is harder and more important. It requires recovering forms of judgment that cannot be automated. It requires asking not only whether language works, but what kind of work it is doing. Is it carrying an earned perception, or merely arranging familiar signals. Does it move by necessity, or only by smooth continuation. Has a mind wrestled with something here, or has language simply kept going until it sounded complete.

These questions are demanding, but they are not optional if we want to remain in contact with serious writing. The alternative is a culture in which language is assessed increasingly by atmosphere alone. That sentence sounds intelligent. This paragraph feels synthetic. That essay appears polished enough to be suspect. Such judgments may occasionally land near the truth, but they are too weak to guide us through what is coming.

A society flooded with generated prose will not be saved by better instinct alone. It will need stronger reading.

It will need readers capable of perceiving that human writing does not announce itself only through warmth or irregularity. Sometimes it appears in the opposite form, in the sentence whose restraint has been earned, in the paragraph whose architecture reflects years of inward discipline, in the prose that does not merely sound intelligent but bears the marks of a person trying, with unusual fidelity, to say exactly what she means.

That fidelity is not machine-like because a machine can imitate some of its surfaces.

The surfaces are not the thing.

The number 74% now seems less offensive to me than symptomatic. It tells me almost nothing about my writing, and rather a great deal about the conceptual confusion surrounding us. We are learning, perhaps too late, that coherence can be counterfeited and that fluency is no longer a reliable sign of mind. But we are also at risk of learning the wrong lesson from that discovery. The wrong lesson would be that polish itself is suspect, or that disciplined prose has somehow become less human because it can be imitated.

The better lesson is more difficult. It is that the distinction we most need was never the distinction between smooth language and rough language. It was the distinction between language that merely functions and language that has been inwardly made.

That distinction still exists.

It exists in seriousness, in renunciation, in proportion, in the felt relation between a sentence and the reality it is trying to bear. It exists in the writer who knows that words are not decorative matter but vehicles of transmission, and who places them accordingly. It exists wherever a person has submitted language to enough discipline that thought can fully appear without excess, blur, or evasion.

If some tools cannot see that, their blindness is unsurprising.

What would be more troubling is if we no longer could.